One Way to Obtain the Smoothed Estimate of Phase Vector in the Task of Motion Control of Convoy Robots

Abstract

In this paper, we formulated and solved the problem of fixed-lag smoothing estimation for continuous linear dynamical systems. The problems of other fixed-lag smoothing algorithm for continuous linear dynamical systems are indicated. Firstly, sometimes the fixed-lag smoothing algorithm is not stable. Secondly, the knowledge of state transition matrix is needed to implement algorithm. It is different to get it, especially for time-vary system. An algorithm is developed for generating the optimal smoothed estimate x(t - T|t) of the state x(t) of a continuous linear system, where t is the most recent measurement and T is a positive real constant. The recurrent algorithm is obtained using the properties of state transition matrix Ф(t 1 , t 2 ) of a linear dynamical system. The developed method makes it possible to bypass the time-consuming procedure of finding state-transition matrix. The process of computing the fixed-lag smoothed estimate is discussed in terms of the algorithm's dependence on the solutions of the filtering and fixed-point smoothing problems. Therefore, the calculation of the estimation of fixed-lag smoothing depends on the value obtained during the filtering process. We give the simulation result of fixed-lag smoothing algorithms for a nonstationary linear system and compared it with the simulation result of filtering algorithms. It is showed that the estimation errors of fixed-lag smoothing are less than the estimation errors of filtering. We also used the fixed-lag smoothing algorithms to improve the localization accuracy of robots in the task of motion control of convoy robots. A node named fix_lag_smoothing is written using C++ in ROS